Here’s Why Talib Kweli and others are Protesting the New Spotify Algorithm
The Imminent Dangers of AI Algorithms Characterizing Peoples’ Interests
The Imminent Dangers of AI Algorithms Characterizing Peoples’ Interests
Recommendation engines are the heart and competitive edge of data-driven companies. It has been one of the most profitable data analytics technologies for companies to increase customer retention and customer addictiveness. A Mckinsey report estimates, 35 percent of what consumers purchase on Amazon and 75 percent of what they watch on Netflix come from product recommendations based on such algorithms.
However, there are also ethical concerns associated with the power of recommendation engines. Cambridge Analytica's involvement with the Trump administration has gained infamy by using algorithms to skew people into watching incorrect, radicalized, extreme videos. Ultimately, playing a pivotal part in the severe polarization of the United States.
Spotify's new algorithm
Not surprisingly, Spotify also invests aggressively in data and analytics. R&D roughly accounted for $1bn in costs in FY2020. Spotify attempts to differentiate themselves through their data and analytics assets, processes, and strategies to maximize monthly active users.
Spotify fully leverages customer data to get the user to listen to more music. The recommendation engine is used to curate the “Discover Weekly” “Daily Mixes” playlists. The playlists include music the users have never heard of before but most probably enjoy on their perfect personal home screen. In addition, “Spotify Radio” pushes music similar to the music you have been wanting. The relentless strive for Spotify to increase customer satisfaction and attention has allowed Spotify’s premium subscribers to grow to over 158 million users.
Spotify’s cutting-edge AI uses three types of information:
・User information: Music history, songs, artists, and genres liked by the user.
・Other users’ data: Other users who have similar tastes in songs, artists, and genres.
・Audio and text information of the song: Automatic categorization of songs into groups that have similar characteristics like mode, rhythm, loudness, and genre of song.
Recently, Spotify has filed patents for some potentially creepy technologies with Access Now (NPO protecting digital rights), and famous musicians nervous about the company’s direction. The new technology will give Spotify the power to change music based on your mood.
The above patent shows how Spotify can use speech recognition to understand your emotional state, gender, age, and accents. Additionally, Spotify can retrieve environmental metadata such as the physical environment (bus, metro, train, etc.) and social environment (alone, small group, party).
The above diagram shows a bird-eye view about the patents ability to combine data from speech recognition with historical listening, user’s taste, friends' taste, and other factors to curate the perfect playlist for you.
On the one hand, this technology may provide the perfect song at the perfect time. If you are at a house party and want to turn up the mood, Spotify can curate the perfect songs to keep the party going. On the other hand, this technology can be seen as emotional manipulation brought to the next level. With the correct recipe and doses of the right songs, Spotify can either aggravate or alter your emotional state. Listening to recommended songs daily can start affecting your personality. Spotify has even conducted a study to show the close relationship between personality traits and the music you listen to.
Concerns raised from Access Now and 180 Musicians
Access Now and 180 Musicians (including Talib Kweli, Tom Morello, Sadie Dupis) mentions that their major concerns with technology are as follows:
Emotion Manipulation: As reinstated above, the technology could potentially give Spotify a dangerous position of power concerning a user
Discrimination: Spotify will try to use speech recognition to infer gender and accent, but this can lead to discrimination, especially against trans, non-binary people, and others who do not fit gender stereotypes.
Privacy Violations: Obviously, continuous monitoring of your emotional state and speech could be seen as a violation of privacy.
Data Security: Risk of leak of personal data collected.
Exacerbate inequality in the music industry: Recommendation engines have the power to push specific musicians and cause an even larger disparity in the music industry.
Given the announcement of Spotify’s technology, Access Now and 180 Musicians have decided to protest the technology publicly. The timeline of events that have transpired since is as below:
April 2, 2021: Access Now sent a letter to Spotify calling on the company to abandon the technology on the patent
May 4, 2021: Access Now reinstates their concerns and asks Spotify to make a public commitment never to use, license, sell, or monetize the recommendation technology by May 18, 2021.
“You can’t rock out when you’re under constant corporate surveillance. Spotify needs to drop this right now and do right by musicians, music fans, and all music workers.” — Tom Morello (guitarist of Rage Against the Machine)
“Instead of wasting money developing creepy surveillance software, Spotify should be focused on paying artists a penny per stream and being more transparent about the data they’re already collecting on all of us.” — Sadie Dupis (Speedy Ortiz and Sad13)
Final Thoughts
Working in the AI Industry, I can vouch that recommendation engines are among the most popular technologies for companies to invest in. One motivation for Spotify behind the patent is to ensure its competitors will not use the same technology. However, once the ethical climate regarding recommendation engines becomes looser, they might slip this feature into their updates one day. Who knows.
If the user enjoys the recommendations and has a positive relationship with the app, is there something wrong? For instance, a “streak” on Duolingo or Simply Piano is seen to be a positive use as it encourages users to maximize their growth.
The toxicity arises when recommendation engines get you to do things you don’t want to do. For example, I don’t want to watch cancerous TikTok videos on YouTube Shorts, but that’s all I seem to be getting on my feed these days.
However, as the user, it is always a choice by comparing the merits and demerits, and I still subscribe to YouTube despite the annoying recommendations. In Spotify’s case, if the alternative is going back to Turntables and CDs, it may be nearly impossible for users to get away from the technology. Thus it is imperative that other stakeholders such as Access Now and musicians lead the cause against the dangerous use of AI and similar technologies.